5 November 2009—Last week, two rods snapped on California's Bay Bridge, raining debris on three vehicles and forcing officials to close the bridge, a major commuter artery, for a week. The inconvenience was minor compared with the August 2007 bridge collapse in Minnesota, which killed 13 people and injured 145, but both incidents have their roots in the United States' aging bridge infrastructure. Hewlett-Packard Laboratories thinks one way to forestall (or at least monitor) the gradual deterioration of the world's bridges is to pepper them with many thousands of small networked acceleration sensors that could, in theory, provide warnings before catastrophic failure.
The company's grand future vision is called Central Nervous System for the Earth (CeNSE), and today HP Labs announced a technological breakthrough that will become a key part of that vision: new inertial sensing technology that will make digital microelectromechanical systems (MEMS) accelerometers up to 1000 times as sensitive as what's currently available. The sensor is based on HP's MEMS technology, which was first commercialized in the company's inkjet printer cartridges.
Though much R&D is focused on creating nanometer-scale parts—turning MEMS into NEMS—that wasn't HP's tack. "There are places where nano is not your friend," says Peter Hartwell, a senior researcher at HP Labs in Palo Alto, Calif. "Inertial sensing is one of those. In the infinite race to build things smaller—to make them cheaper and lower power and so on—in inertial sensing it actually hurts you."
Until now, the paradigm has been that engineers could build high-end expensive accelerometers (like those used in aircraft guidance and seismic monitoring), or they could build low-end, less-exact consumer-grade accelerometers on the cheap—like those that populate the Wii remote or trigger the air bags in your car. These were the fundamental trade-offs between sensitivity and cost.
But HP has rethought the way MEMS inertial sensors are built, says David Erickson, an engineering manager at HP's Technology Development Organization, making it possible to produce a cheap but sensitive device. They did it by changing the accelerometer's innards, specifically the "proof mass" [see illustration], which is a piece of silicon suspended on a stiff silicon spring or flexion inside the MEMS device that moves when the sensor moves. Electrodes on the mass, and in an unmoving section of silicon such as the device's surface, form a capacitor. The capacitance changes when the mass moves, and that change translates to acceleration information.
It turns out that there is a limit to the smallest measurable acceleration signal (called a noise floor), and that limit is set by the thermal vibration of the atoms in the proof mass. In a small proof mass, statistics take over. Taking a simple model as an example, if the mass contains 1000 atoms, 50.1 percent of atoms might jump one way, and 49.9 percent might jump the other way. As a result, the proof mass moves. "We can measure the position of the proof mass accurately enough to see it wiggling from that thermal energy," says Erickson. If the proof mass is too small, those vibrations are indistinguishable from acceleration.
To solve that, Hartwell says, you need to add mass. With a big enough mass, the actions of the atoms are more likely to cancel one another out. The larger the number of atoms in the proof mass, "the smaller that [thermal] vibration," he says. The proof mass inside HP's new device is 1000 times as massive as the proof mass inside most consumer MEMS accelerometers. "So that dramatically improves the sensitivity of the device," says Hartwell—by a factor of 1000.
Making a component chunkier might seem easy, but in fact, most MEMS production processes can't do it. The new device has the sensitivity needed for high-end applications, but it won't carry the old price tag. The sensing device will not be available for individual sale. Rather, HP wants to piggyback the technology on other sensors, so the accelerometer is essentially free.
HP envisions 1 trillion sensors in use around the world, creating a central nervous system in the form of a complex, far-flung sensor network that could monitor climate change, help with oil and gas discovery and seismic monitoring, and likely be useful in monitoring the health of the United States' roughly 600 000 bridges (over one-fourth of which are considered "deficient" or "obsolete," according to the American Society of Civil Engineers). For bridges specifically, Hartwell envisions a final CeNSE sensor roughly the size of a thumbtack "that you can literally stick into a bridge." A larger bridge would require about 10 000 sensors to form the network, he says, while a smaller overpass might require a few hundred.
But for now, the sensors will be about the size of a cigarette pack. The reason is power. Power has been a critical problem for high-performance sensors, which require about 1 watt. Conversely, MEMS for consumer devices like those in a Wii-mote require tens to hundreds of milliwatts. HP's new sensor is a big improvement, because while it increases performance to the level of those high-performance sensors, it requires less than 50 milliwatts.
But before even the cigarette-pack-size low-power sensors can become a reality, the cost of sensors will need to come down dramatically. "To fulfill our CeNSE vision, the sensors really have to be just about free," says Hartwell, "because we want a trillion of them. So they'll have to be integrated with other kinds of devices."